SUMMARY: Problem with mirrors on Ultra2

Pascal Gaudette (pascal@viper.istar.ca)
Thu, 20 Mar 1997 10:14:09 -0500 (EST)

Last week I posted about problems I've been having in trying to
boot an Ultra 2 with mirrored partitions with a single disk
(contingency testing). My original post is included below, but
basically I can get the machine to boot if I pull out the drive
at SCSI ID 0, but not if I pull out the one at SCSI ID 1.

I received some good suggestions, but unfortunately none of them
actually solved the problem. One person suggested that since I'm using
DiskSuite 4.0, there's a big patch (102580) that fixes many bugs. I
applied it, but the problem remains. Someone else suggested there may
be problems with the meta device state replicas, but that turned out
not to be the case. So the problem remains. Hopefully this situation
won't present itself in non-testing, real-life operations
(dreamdreamdream)...

Thanks to John Malick <john@starinc.com> and Matthew Stier
<mstier@hotmail.com> for their help.

--
Pascal Gaudette <pascal.gaudette@corp.istar.ca>
UNIX Programmer/Analyst, iSTAR internet inc.
#include <std/disclaimer.h>




--------8<--------8<--------8<--------8<--------8<--------8<--------
Date: Wed, 12 Mar 1997 18:34:00 -0500 (EST)
From: Pascal Gaudette <pascal@viper.istar.ca>
To: sun-managers@ra.mcs.anl.gov
Subject: Problem with mirrors on Ultra2 - stumped


Hi folks,


Well, I'm stumped. I've been doing a few tests on an Ultra 2 running
Solaris 2.5 with 2 2GB drives at SCSI IDs 0 and 1, partitioned
identically and set up with DiskSuite4.0 to mirror everything but swap
(which is split evenly accross the two disks). And I've run into a
little problem.


My tests consist of bringing the machine down (including power off),
pulling out one of the drives, booting and seeing if I can get to a
functional state. (There's a second part if I get to a functional
state, but that need not be of concern since the problem is in the
first part.) Note that the machine runs fine with both drives in it,
with metastat reporting every mirror OK and metadb -i reporting
everything fine (3 copies of the state DB on a small partition on each
drive).


If I pull out the drive at SCSI ID 0, everything works out. I can get
to a fully functional state (it's a bit of an involved process, but it
works). But if I remove the drive at SCSI ID 1, I get:


> Cannot mount root on /pseudo/md@0:0,50,blk fstype ufs
> panic[cpu0]/thread=0x10402040: vfs_mountroot: cannot mount root


And it dies. Can anyone help me diagnose the problem? 




Here are some relevant file portions:


/etc/system:
> rootdev:/pseudo/md@0:0,50,blk
>


This seems like the likely culprit. But it was automatically set up
using "metaroot /dev/md/dsk/d50", so I don't really know what to
do with it...


Root partition:
> # ls -l /dev/dsk/c0t0d0s0
> lrwxrwxrwx   1 root  root  51  Jan 17 15:26 /dev/dsk/c0t0d0s0 -> 
> ../../devices/sbus@1f,0/SUNW,fas@e,8800000/sd@0,0:a


/etc/vfstab:
> #device         device            mount   FS      fsck    mount   mount
> #to mount       to fsck           point   type    pass    at boot options
> /dev/md/dsk/d50 /dev/md/rdsk/d50  /       ufs     1       no      -


/etc/opt/SUNWmd/md.tab:
> # Root partition
> d0 1 1 /dev/dsk/c0t0d0s0
> d1 1 1 /dev/dsk/c0t1d0s0
> # Metamirror for /
> d50 -m /dev/md/dsk/d0 /dev/md/dsk/d1


Any help would be greatly appreciated. I will summarize answers to
the list.



--
Pascal Gaudette <pascal.gaudette@corp.istar.ca>
UNIX Programmer/Analyst, iSTAR internet inc.
#include <std/disclaimer.h>